Short-time Forecasting And Rolling Correction Methods For Global Radiation

Solar energy is a clean energy, with rational development of which will reduce air pollution and help protect the environment. It is also of great meaning in dealing with climate change and increasing the safety of energy consumption. In order to achieve the sustainable development of energy and environment, solar energy development has been treated as one of the primal goal in many countries lately. Therefore the solar energy has became the third renewable energy next to hydraulic power and wind power. With the rapid development of solar energy, the requirement of radiative forecast has increased. Many forecasts have been given by using mature mesoscale numerical models with high resolution, like MM5 and WRF. But the prediction error of radiative forecasts under cloudy and rainy conditions is still very high.The hourly observational data of total radiation from the southern sub-urban observatory of Beijing and the hourly WRF model simulated data containing 14 conventional meteorological elements for the next 24 hours has been used in this work to testify a variety of both linear and nonlinear methods in correcting the simulated total radiation. The differences of corrective accuracy between methods on seasons and the differences in sample numbers have been discussed in detail, and the optimal rolling correction scheme has been determined through this work. The results are as below. Firstly, the weights of influence factors were different through seasons. However, the forecasting radiation, the specific humidity and the relative humidity at 2 meters above the ground, low and middle cloud cover, the mixing ratios of water in different phases in cloud and the atmospheric water vapor mixing ratio were the recommended factors in the total radiation correction. Secondly, the optimal sample number must be selected according to season. Thirdly, hourly rolling schemes were much higher in corrective accuracy than non-rolling schemes. By applying the hourly rolling schemes, the average absolute error/ mean absolute percentage error/ relative root mean square error of the simulated total radiation were reduced by about 55.76 W·m-2/ 19.01%/ 26.38%, and 79% of results were improved in absolute error compared to uncorrected results. Furthermore, the absolute errors of corrected total radiation were presented in typical normal distribution pattern, mostly concentrated in between-80-40 W·m-2, and the maximum distribution range was form -40 to 0 W·m-2. The radiative corrections were better in autumn and winter then those in spring and summer, and it is caused by the differences in the forecasting of the macro and micro physical parameters of cloud among...